Splice variants of DOMINO control Drosophila circadian behavior and pacemaker neuron maintenance

Mass spectrometry (MS) label-free quantitative proteomics approach was previously performed to identify the BRAHMA (BRM) chromatin-remodeling protein complex as interactor of circadian clock proteins [31]. In the same data set, we observed that several core subunits of the ATP-dependent DOM chromatin-remodeling complex, including DOM, NIPPED-A, PONT, REPT, and Mrg15, interact with CLK in the nucleus of Drosophila S2 cells (Table 1). Since DOM is the ATPase subunit of the protein complex and shows significant binding to CLK (especially C-terminal FLAG-tagged CLK) according to SAINT (Significant Analysis of INTeractome) scoring [32], we decided to further investigate its role in circadian regulation. Interestingly, prior studies indicated that dom mRNA levels in the sLN_vs are clearly enriched about 4.4-fold compared to other neurons [33].

Homozygous dom mutants are lethal, therefore we used RNAi to deplete DOM in circadian neurons. Dicer 2 was co-expressed to enhance RNAi efficiency and has been shown to have no effects on circadian rhythms [34, 35]. When we expressed dom dsRNAs using tim-GAL4, a circadian tissue-specific driver [36], most of the flies became arrhythmic, and the amplitude of rhythm was significantly reduced (Fig 1A and S1 Table). Only 31% of flies with dom downregulation (41% for another independent line dom RNAi#2) were rhythmic, and interestingly the circadian period of these lines was about 1.5 hours longer than the control. The reduced rhythmicity and amplitude were also observed when we only targeted the PDF positive LN_vs using pdf-GAL4 [11] (Fig 1A). However, no reduction on circadian rhythmicity was observed when we targeted all PDF negative circadian cells, using a combination of tim-GAL4 with the repressor transgene pdf-GAL80 (Fig 1A and S1 Table). Only a slight period-lengthening of activity rhythm (~0.4hr) was detected. These results indicated that DOM primarily functions in the LN_vs to control circadian behavior, but could function in PDF negative circadian neurons to fine-tune circadian period length.

Because of the high arrhythmia in DD, we also examined the behavioral phenotype of dom downregulation in tim-expressing circadian neurons under LD. We found that DOM knockdown abolished the morning peak of anticipatory activity in flies (Fig 1B). Importantly, two independent dom RNAi lines targeting non-overlapped regions of dom exhibited similar phenotypes in both LD and DD (Fig 1C and S1 Fig). Thus, it is unlikely that off-target effect of dom RNAi cause these circadian phenotypes. Furthermore, quantitative RT-PCR results showed that dom expression were significantly reduced in fly heads of DOM knockdown as compared to controls (S1 Fig). Together, these results suggest that DOM is important for regulation of circadian locomotor behavior.

Depletion of DOM severely decreased the rhythmicity and lengthened circadian period. To understand how DOM affects circadian rhythms, we first measured the oscillations of mRNA abundance of three core pacemaker genes in fly heads with expression of dom dsRNAs using tim-GAL4. For clk, the abundance and oscillation of mRNA under LD and DD were not affected in dom RNAi flies (Fig 2A). However, we found that the abundance of per and tim mRNA was reduced, especially at the time point of peak expression under LD (Fig 2A). Under DD, depletion of DOM also decreased the level of per and tim mRNA, which indicates that the effects are not due to masking effects of light (Fig 2A). We hypothesized that DOM may facilitate CLK-driven rhythmic transcription by regulating the balance between H2Av and phosphor-H2Av via binding to CLK. So we examined the H2Av occupancy at per and tim promoter by performing chromatin immunoprecipitation (ChIP) in fly heads using the commercialize Drosophila H2Av antibodies. We entrained flies under LD and collected head samples at ZT4 and ZT16, close to the trough and peak time of CLK binding [37]. Consistent with our hypothesis, we observed a significant increase in H2Av binding at both promoters when dom is knocked down in tim-expressing circadian cells after normalized to an intergenic region control (Fig 2B). This data indicate that DOM interacts with CLK to regulate transcription of circadian genes by replacing H2Av at the promoters. However, we did not observe significant difference in binding between ZT4 and ZT16 (Fig 2B), which indicates that the effect of DOM might not be time-dependent.

Next we examined the oscillation of PER protein in major groups of pacemaker neurons with tim-GAL4 expressing dom dsRNA across 24 hours under DD. Consistent with the changes at mRNA level, the level of PER was also significantly reduced in the sLN_vs with DOM depletion (Fig 2C and 2D). In other groups of circadian neurons, such as the LNds and DN1s, decreased abundance of PER was also observed (Figs 2D and S2). Taken together, our results suggest that DOM regulates the H2Av occupancy at promoters of circadian genes and controls the abundance of PER.

There are two major splice variants of dom: dom-A and dom-B, which has distinct functions during Drosophila oogenesis [25]. We wondered whether these two isoforms would have different functions in circadian rhythms. We first examined the expression pattern of dom-A and dom-B in the fly head across different times of the day. Interestingly, dom-A exhibited a strong oscillation pattern with a trough around zeitgeber time 9 (ZT9, as ZT0 is light on and ZT12 is light off) and a peak expression near ZT21 (S3A Fig). Oscillation of dom-A was also observed under DD. In addition, the oscillation of dom-A expression was abolished in the Clock null mutant Clk^out (S3A Fig) [38]. Together, these data indicate that dom-A is controlled by circadian clock. However, we did not observe a clear oscillation for dom-B expression (S3B Fig). In order to detect changes at protein level, we took advantage of isoform-specific antibodies to DOM-A or DOM-B [25]. We confirmed the specificity of these antibodies by altering the level of DOM-A or DOM-B using RNAi and overexpression (S3C and S3D Fig). Similar to mRNA level, DOM-A protein abundance was dramatically reduced in Clk^out, however, no oscillation of DOM-A was observed in wild type flies (S3E and S3F Fig). This loss of DOM-A oscillation at protein level indicates that DOM-A protein might be quite stable. The levels of DOM-B were not affected in Clk^out (S3G and S3H Fig).

Differential regulations of DOM-A and DOM-B by CLK suggest that they might have different roles in circadian rhythm. We therefore performed isoform-specific downregulation in circadian tissues using small hairpin RNA (shRNA) targeting dom-A or dom-B. These transgenic fly lines have been previously shown to specifically knockdown these alternative isoforms (S1 Fig) [25]. We first quantified the efficiency of dom-A and dom-B downregulation by expressing shRNAs in circadian cells of fly heads using quantitative real-time PCR. Consistent with previous report [25], we observed that each shRNA line specifically downregulated dom-A or dom-B (S1B Fig). Interestingly, when we depleted DOM-A or DOM-B in all circadian clock neurons by tim-GAL4, we observed two distinct circadian behavioral phenotypes. Most of the flies expressing dom-A shRNA lost locomotor rhythms, except for ~20% that were rhythmic and showed a longer period as compared to the control (Fig 3A and 3B; S1 Table). As with depletion of all dom isoforms, DOM-A specific knockdown also blunted the morning activity peak under LD (S4 Fig). However, for DOM-B knockdown, flies exhibited a lengthened period of ~1.5 hrs in DD (Fig 3A and 3B), but no effect on amplitude of rhythms or morning anticipation was observed under LD (Fig 3B and S4 Fig). When we expressed dom-A or dom-B shRNA only in LN_vs using pdf-GAL4, or in PDF negative circadian neurons, we found similar effects on rhythmicity or period-lengthening (Fig 3B). These results suggested that DOM-A and DOM-B are required in both PDF positive and negative circadian neurons for behavior. It is unlikely that the different circadian phenotype is due to greater efficiency of dom-A over dom-B shRNA (S1B Fig). Actually a previous study has also validated the knockdown efficiency and found that dom-B shRNA is in fact slightly stronger than dom-A in the fly nervous system [25]. However, we cannot fully exclude the possibility that DOM-A protein is more stable than DOM-B and cause this difference of circadian phenotypes.

Our isoform-specific knockdown indicated that DOM-A and DOM-B have distinct functions in circadian rhythms. Although the shRNA lines have been shown here and in a previous study [25] to specifically knockdown the dom-A or dom-B isoforms, off-target effects could still contribute to the observed phenotypes. Thus, we performed rescue experiments with UAS transgenic flies expressing either dom-A or dom-B cDNA. Circadian behavior defects were rescued when we overexpressed corresponding UAS lines in the DOM-A or DOM-B knockdown (Fig 3A and 3C), which indicated that the isoform specific phenotype we observed was not due to off-target effects. Remarkably, with overexpression of dom-A, we were not able to rescue the long period phenotype of DOM-B knockdown (Fig 3C, and S1 Table). Similarly, overexpression of dom-B could not rescue the arrhythmic phenotype of DOM-A knockdown, although we did notice that the period-lengthening effect was partially rescued in the remaining 13% of rhythmic flies (Fig 3C).

Since we observed DOM regulates the occupancy of H2Av at per and tim promoters, we therefore examined whether DOM-A and DOM-B might be bound to the promoters of per and tim. We used the DOM-A- and DOM-B-specific antibodies to perform ChIP in fly heads with overexpression of dom-A or dom-B at two time points: ZT4 and ZT16. We observed significant enrichments of DOM-A and DOM-B binding on both per and tim promoters, compared to an intergenic region control (Fig 3D). Similar with H2Av ChIP, we did not observe significant difference in binding between ZT4 and ZT16 (Fig 3D), which indicates that the binding of DOM-A and DOM-B might not be time-dependent.

The blunted morning activity peak and arrhythmia in DOM and DOM-A knockdown flies suggest that there might be defects in the sLN_vs or in PDF signaling. There are three major projections of PDF positive LN_vs. The lLN_vs send projections to the optic lobes and the contralateral brain hemisphere, while the sLN_vs send projections to the DN1s and DN2s region [39]. Based on whole mount immunohistochemistry in fly brains, we did not observe obvious defects in the brain structure or projections of the lLN_vs, however, PDF levels in the dorsal projections of the sLN_vs were barely detectable in DOM and DOM-A knockdown flies (Fig 4A). Absence of the dorsal projection or low PDF expression in the sLN_vs could lead to decrease of PDF in the dorsal projection [20, 40]. Our data are consistent with both possibilities. Using pdf-GAL4 and a membrane tethered GFP (CD8-GFP) to label axons of sLN_vs, we observed that the dorsal projections of sLN_vs were clearly shortened in DOM and DOM-A (Fig 4B–4D). Interestingly, the sLN_vs dorsal projections were normal in DOM-B downregulation, which is consistent with the behavior results. Furthermore, close observation of the PDF positive sLN_vs revealed that both the number of sLN_vs and PDF levels were reduced (Fig 5A and 5B). For each hemisphere, the average number of PDF positive sLN_vs in DOM and DOM-A knockdown flies was approximately 2, compared to 4 in the control flies or DOM-B knockdown (Fig 5A and 5B). However, with DOM-B knockdown, we did not observe any obvious loss of sLN_vs or reduction in PDF levels, which indicates that this process is specifically controlled by DOM-A (Fig 5A and 5B). Consistent with the ChIP results, the levels of PER and TIM were found to be significantly decreased in the sLN_vs with knockdown of DOM-A or DOM-B (Fig 5A and 5B). This may also explain why depletion of DOM-A or DOM-B had an effect on period-lengthening.

Next we examined which step of regulation causes the reduction of PDF levels in DOM and DOM-A knockdown. DOM regulates gene expression at the transcription level [24]. We therefore examined the expression of pdf using a transcriptional reporter line, pdfTomato [21]. Consistent with PDF staining, we observed significant reduction of pdf transcription in sLN_vs of both DOM and DOM-A knockdown (Fig 5C and 5D). Interestingly, the decrease in neuron number and reduction of PDF level were unique to sLN_vs, which were not detected in the lLN_vs (Fig 5C and 5D).

In summary, consistent with differences in circadian behavior, DOM-A and DOM-B knockdown also exhibit differences at the molecular level. Knockdown of DOM-A causes loss of sLN_vs and decrease in sLN_v PDF levels, which is not observed in DOM-B downregulation.

Decrease of sLN_vs numbers and dorsal projections in adult flies might be due to abnormal development or maintenance defects. Therefore, we performed brain dissections at larval stage and pupal stage. The number of sLN_vs precursor in the DOM knockdown condition was unchanged compared to the control at 3^rd instar larvae (Fig 6A and 6B), which suggests that DOM does not affect sLN_vs development. However, same as adulthood, even at early pupal stage (3 days after pupation), there were only 1–2 sLN_vs detected in dom downregulation (Fig 6C and 6D). These data indicate that DOM is required for the maintenance but not for the development of sLN_vs. DOM probably affects sLN_vs during larval-pupal metamorphosis. Consistent with adult flies, both PDF and PER levels of dom downregulation were reduced in sLN_vs compared to the control in larvae or pupae (Fig 6B and 6D).

The decrease of axonal projections and number of sLN_vs we observed in DOM knockdown flies (Figs 4 and 5) suggested that DOM might play a role during development to properly regulate circadian rhythms. Thus, we used tim-GAL4 and GAL80^ts TARGET system to temporally express dom dsRNA during development or after eclosion [41]. GAL80 is active at 18°C thus blocks GAL4 function and prevents dsRNA production, while 30°C inactivates GAL80 and allows expression of dsRNA to downregulate DOM. When we grew flies at 18°C and tested circadian behavior at 30°C, we observed strong rhythmicity but around 1–2 hr period-lengthening with adult-specific DOM or DOM-A knockdown (Fig 6E and S1 Table). However, knockdown DOM-A only during development led to high arrhythmic activity (Fig 6E and S1 Table). Thus, developmental expression of dom or dom-A dsRNA appears to affect the rhythmicity, while adult specific depletion appears to be sufficient for period-lengthening. Furthermore, knockdown of DOM-B only in adulthood caused a ~1 hr increase in period length, suggesting that unlike DOM-A, the DOM-B splice form is mainly necessary for the regulation of circadian clocks in adult stage (Fig 6E and S1 Table). We observed a slight but significant lengthening of period (~0.6 hr, S1 Table), when we specifically depleted DOM-B in circadian neurons during development, which indicates that DOM-B might also play a role during development.

To exclude anatomical defects in the circadian circuitry, we grew flies at 18°C and entrain them under LD for 5 days at 30°C, then performed brain dissections at ZT1. As expected, the dorsal projections and number of sLN_vs in the flies with adult-specific DOM downregulation were normal, and PDF abundance was not affected (Fig 6F and 6G). These data further suggest that the arrhythmia we observed in DOM knockdown was mainly due to the function of DOM during development. Indeed, the abundance of PER in sLN_vs was still reduced, which could explain the lengthened period phenotype (Fig 6G). Since we observed DOM binding to the per and tim promoters, and since DOM is required at adulthood for the period-lengthening, we checked whether adulthood specific downregulation of DOM affects per and tim transcription by using quantitative RT-PCR. The pattern of per and tim oscillation was still present in adult-specific DOM knockdown, however the abundance of transcripts is dramatically reduced, especially for the early night time points of per (S5 Fig), which is consistent with the PER protein level (Fig 6F and 6G).

Taken together, these data indicate that DOM/DOM-A is required during development for the rhythmic activity, while both DOM-A and DOM-B are necessary for period determination post development.

Since we observed low PDF levels in the sLN_vs with DOM knockdown, we wondered whether the decrease of PDF signaling was the cause of arrhythmia. To test this hypothesis, we determined whether hyperactivation of PDFR could rescue the phenotypes that accompany reduction in DOM expression. To increase PDFR signaling, we expressed a membrane-tethered PDF (t-PDF) in circadian cells with tim-GAL4, which has been shown to mimic high PDF levels [42]. A scrambled peptide sequence (t-PDF SCR) with similar length as PDF was used as a negative control. Strikingly, co-expression of t-PDF using tim-GAL4 restored the rhythmicity to 89% for dom RNAi (71% for another independent dom RNAi line)(Fig 7A and 7B), whereas the t-PDF SCR control was unable to restore rhythmicity (Fig 7A and 7B). Not only the rhythmicity, but also the period of DOM knockdown was rescued with co-expression of t-PDF (Fig 7B). To exclude the possibility that the behavior rescue in t-PDF was due to ceased expression of dom dsRNA, we measured the dom transcripts in fly heads. Co-expression of t-PDF with tim-GAL4 downregulated dom to the similar level as dom RNAi control (S6A Fig). Since the circadian period was also restored in dom knockdown with expression of t-PDF, we then stained fly brains with PDF and pacemaker proteins PER and TIM (Fig 7C and S6B Fig). With expression of t-PDF in DOM knockdown flies, we made two interesting observations. First, PDF level as well as the dorsal projection of sLN_vs were restored, while the number of sLN_vs was still lower (Fig 7C and 7D). Second, the abundance of PER and TIM in sLN_vs was also restored (Fig 7C and 7D and S6B and S6C Fig). Together, our data showed that activation of PDFR signaling in circadian neurons rescues the arrhythmic and period-lengthening phenotype of dom knockdown.

All the flies were raised on cornmeal/agar medium at 25°C under a LD cycle. The following strains were used: yw,w¹¹¹⁸, yw; tim-GAL4/CyO [36], yw;Pdf-GAL4/CyO [11]; yw;tim-GAL4/CyO;Pdf-GAL80/TM6B [47], tublin-GAL80^ts, UAS-cd8-GFP, clk^out. The following stocks were obtained from the Bloomington Drosophila Stock Center (http://flystocks.bio.indiana.edu/↗): UAS-dom^RNAi#1 (BL31054), UAS-dom^RNAi#2 (BL38941), UAS-domA (BL64261) and UAS-domB (BL64263). DomA^RNAi(sh-domA) and domB^RNAi(sh-domB) fly lines were generous gifts from Dr. Peter B. Becker. PdfTomato line was generated by Dr. Sebastian Kadener.

For behavioral experiments, adult male flies (2–4 days old) were used for testing locomotor activity rhythms. Flies were entrained for 4 full days LD cycle at 25°C, using about 500 lux light intensities, and then released into DD at 25°C for at least 6 days. Locomotor activity was measured with TriKinetics Activity Monitors in I36-LL Percival Incubators. Locomotor activity was averaged over the 4 days entrainment for LD and 6 days for DD. Analyses of period and power were carried out using FaasX software as previously described [48]. Actograms were generated using a signal-processing toolbox implemented in MATLAB (MathWorks) [49]. For GAL80^ts experiments, flies were raised at 18°C and tested at 30°C. They were entrained for 5 days and then released in DD for at least 6 days. Morning anticipation was calculated by the ratio of activity counts between 2 hrs before light on and 6 hrs before light on. We first measured the single fly activity counts obtained in twelve 30-min bins between Zeitgeber Time (ZT) 17.5 and ZT24 (6 hr before lights on) and six 30-min bins between ZT22.5 and ZT24 (2 hrs before lights on). The first value is divided by the second to obtain the morning activity index. Morning anticipations of individual flies were then averaged and plotted on the graphs.

Whole-mount immunohistochemistry for fly brains were done as previously described [50]. Adult fly (3–6 days old) or L3 stage larval brains were dissected in chilled PBT (PBS with 0.1% Triton X-100) at the indicated time points and fixed in 4% formaldehyde diluted in PBS for 30 min at room temperature. For pupal brains dissection, Drosophila were fully developed and hatched 7 days after pupation in the same cross, pupa were collected 3 days after pupation using a small wet paintbrush and transfer to a dissection dish. After 2–3 times PBS wash, the pupal brains were dissected and fixed as previously described method. The brains were rinsed and washed with PBT three times (10 min each). Then, brains were incubated with 10% normal Goat serum diluted in PBT to block for 60 min at room temperature and incubated with primary antibodies at 4°C overnight. For PER, TIM and CLK staining, we used 1:1,500 rabbit anti-PER, 1:2,000 rat anti-TIM (gift from Dr. Rosbash) and guinea pig 1:2,500 anti-CLK (gift from Dr. Hardin), respectively. We used a 1:200 dilutions for mouse anti-PDF and 1:200 for rabbit anti-GFP (DSHB). After six washes with PBT (20 min each), brains were incubated with relative secondary antibody at 4°C overnight, followed by another six washes with PBT. For PdfTomato staining, flies were directly dissected in chilled PBT at the indicated time points and moutained in the medium (VECTOR). All samples were imaged on a Leica Confocal SP8 system, with laser settings kept constant within each experiment. 10 to 12 fly brains for each genotype were dissected for imaging. Representative images are shown. Image-J software (National Institutes of Health [NIH]) was used for PER, PDF, TIM and CLK quantification in 15–30 sLNvs from at least seven brains. For quantification, signal intensity in each sLNv and average signals in three neighboring non-circadian neurons were measured, and the ratio between signals in sLNvs and non-circadian neurons was calculated.

Chromatin immunoprecipitation (ChIP) was performed based on published protocols [31]. Flies entrained in 12 hr light:12 hr dark (LD) conditions at 25°C for four days were collected at two time-points (ZT4 and ZT16) on the fifth day. Briefly, chromatin was isolated from 500 μl of fly heads homogenized with a dounce homogenizer (Kimble Chase) for 20 strokes using the loose “A” pestle. Homogenate was sieved by a 70 μm cell strainer (Falcon) then centrifuged to remove cell debris. Pellets were cross-linked using formaldehyde. Samples were sonicated using a focused-ultrasonicator (Covaris M220) on setting for 400–500 bp cDNA and then centrifuged at 10,000 rpm for 10 minutes. Supernatant was collected in two 130 μl aliquots for IP and 26 μl was collected for input and frozen at -80C for analysis. Sonicated chromatin was roughly 500 bp in length (<1000 bp). For each IP, 30 μl of a Protein G Dynabead slurry (Life Technologies) was washed then incubated along with the appropriate antibodies for 4 hours at 4°C with rotation. Amount of antibodies used for ChIP is as follows: anti-H2Av (10 ug/ml, rabbit, Active Motif), anti-DOM-A (20 μg/ml, rabbit, GenScript) which was generated in our lab (antigen protein sequence designed according to 2008–2349 aa of DOM-A), anti-rabbit-IgG (20 μg/ml, Life Technologies), anti-DOM-B (10 μg/ml, mouse, from Dr. Peter B. Becker), anti-mouse-IgG (10 μg/ml, Life Technologies). Following incubation, beads were collected and incubated with chromatin overnight at 4°C with rotation. DNA was eluted using the Qiagen PCR purification kit and subjected to qPCR. At least three technical replicates of qPCR were performed for each biological ChIP replicate and three biological replicates were performed for H2Av, DOM-A and DOM-B assays. Background binding to a nonspecific antibody (anti-IgG; Life Technologies) bound to Dynabeads was subtracted from input samples and results are presented as the percentage of the input samples. For each assay, at least three biological replicates were performed. The specific primers used for qPCR are described in S2 Table. The technical qPCR triplicates were averaged for each biological replicate as no significant differences were found between the technical replicates, and the error bars represent SEM calculated from variance between biological replicates. Two-tailed t-tests were used to determine statistical differences between control and experimental treatment at each ZT.

Flies were collected at the indicated time points and isolated heads were stored at −80°C. Total RNA was extracted from 25–30 heads with TRIzol based on the manufactures protocol (Life Technologies, USA). A 2-μg quantity of RNA was reverse-transcribed with reverse transcription reagents (Invitrogen). For real-time PCR (qRT-PCR) of per, tim, clk, dom, domA, domB and actin, we used a qPCR detection kit (SYBR Select Master Mix For CFX) (Life Technologies, USA). The specific primers used for qPCR are described in S2 Table. All the experiments were performed in the CFX96 Real-Time System (BIO-RAD).

Fly heads were collected at the indicated time points and homogenized with pestles, protein extracts were prepared with HEPES-Triton lysis buffer (1X HEPES-Triton buffer, 1 mM DTT, 0.4%NP-40, 0.1% SDS, 10% glycerol, 1X tablet protease inhibitor). Proteins were quantified using BSA assay. For immunoblot analysis, proteins were transferred to PVDF membranes (Genesee Scientific) and incubated with anti-DOM-A (1:300, GenScript), anti-DOM-B (1:5, from Dr. Peter B. Becker) and anti-ACTIN (1:100,DSHB) in blocking solution. Band intensity was calculated and analysed with Image J.

Statistical analysis of two data points was performed with Student’s t-test. Statistical analysis of multiple data points was performed with one-way analysis of variance with Tukey post-hoc tests using GraphPad software.